Our research has not found consistent health declines following retirement. This continuation project hypothesizes that maintenance of the quantity and quality of social supports reduces retirement stress so that anticipated health declines are averted. The stress and health literature tends to focus on global measures of stressful events, on mental health or on indices of physical health such as physician visits or self-reports, and consists primarily of cross-sectional designs. The dependent variable in this continuation project will be change in physical health. This will provide a more rigorous test of the support-health hypothesis than would the use of mental health measures which risk circularity in perception of support and perception of distress. The independent variable will be change in social supports from before to after the single stressful event f retirement. Social support will be measured by the Social Relationship Scale which provides a measure of the quantity and quality of supports. Our prospective design will be based on new data to be collected and on existing longitudinal data. Participants in the study, aged 55-74, will be drawn from the 1900 male volunteers of the Normative Aging Study, a longitudinal study of aging in initially healthy males conducted since 1963 at the VA Outpatient Clinic, Boston. New social support data will be collected in the first year of this 5-year project when participants (N=180) are still working and again through the next three years as participants pass from work to retirement. Pre- and post-retirement health data will be obtained from the medical examinations administered by the core Normative Aging Study project every three years. Existing data (N=489), collected prospectively between 1975-1981, are limited to social support questions on work and retirement answered when respondents were working and again after they retired. These existing pre- and post-retirement data cover a maximum span of up to six years into retirement. Data analyses will primarily employ regression analysis using residualized scores of change in social supports and change in health. The long-term objectives of this study are to understand conditions which promote the maintenance of health among retirees and thereby help guide programmatic or social interventions on behalf of older people. Such knowledge will help health practitioners identify patients who are at greater risk because of deficiencies in their social networks.